摘要
传统推荐技术存在冷启动、稀疏性、推荐精度低等问题。其中，可以方便表达用户兴趣偏好的标签推荐存在噪声、一词多义等问题，稳定性较好的用户兴趣刚好可以解决这一问题。然而，在推荐技术领域内，将兴趣与标签相结合的推荐研究相对较少。本文提出基于兴趣-标签的推荐算法ITRA（Interest—Tag Recommendation Algorithm），通过定义计算用户兴趣权重值、用户兴趣相似度、用户候选兴趣集、推荐兴趣-标签集、项目推荐集，将该集合作为最终的推荐结果。最后，通过实验证明该算法可以有效的提高推荐结果的准确率。
Traditional recommendation technologies have many disadvantages, such as cold start, sparseness as well as low recommendation accuracy. Among these technologies, tag recommendation can express users＇ interests very well, however it still exists some problems such as noise interference and polysemy. In such situation, users＇ interests are more stable and can be used to solve the problems mentioned above. While only several studies have combined interest and tags in recommendation area. This paper put forward ITRA （Interest-Tag Recommendation Algorithm ） to deal with the condition. ITRA was able to calculate the weight of users＇ interests and the similarities between users＇ interests. On this basis, it could get the candidate set of user interest together with the recommendation set of interest-tag, and in the end recommended the items set to users. Finally, the experimental study can verify the improvement of recommendation accuracy by using this algorithm.
出处
《情报学报》
CSSCI
北大核心
2015年第5期466-470,共5页
Journal of the China Society for Scientific and Technical Information
基金
国家科技支撑计划(2013BAH13F01)
关键词
个性化推荐
用户兴趣
兴趣标签
电子商务
personalized recommendations, user interest, interest tag, e-commerce
作者简介
李兴华，男，1990年生，武汉理工大学电子商务研究生（1134099456@qq．com）；
陈冬林，男，1970年生，博士生导师，主要研究方向：云计算、服务管理、商务智能；
杨爱民，男，1970年生，讲师，主要研究方向：智能推荐、企业资源计划（ERP）、企业间供应链集成．
